广域级电网在运智能电能表的烧损故障关联分析和预测方法

针对智能电能表在运行过程中出现烧损的现象，在对各类因素进行关联分析后， 提出了一种基于XGBoost算法的智能电能表烧损预测方法，以某省份2019—2020年的数据为例进行了测试验证. 采用该方法结合电能表基本信息数据、运行数据和环境数据进行烧损识别，并与K最邻近法（K-NearestNeighbor，KNN）、朴素贝叶斯和支持向量机等传统机器学习算法进行对比. 结果表明，基于极限梯度提升算法（eXtreme Gradient Boosting， XGBoost）的算法精度达到91%，召回率达到66%，综合指标F1达到76.51%，远高于传统算法. 算法模型在进行系统部署的过程中，运用长短期记忆算法（Long Short Term Memory， LSTM）对部分缺失值进行了填充，经试点验证，该模型可较为准确地预测低压台区电能表烧损现象.

Burning Fault Correlation Analysis and Prediction of Smart Meters in Operation in Wide-area Power Grid

Aiming at the burning loss of smart electricity meters, after correlation analysis of various factors, this paper proposes a burning fault prediction for smart meter based on XGBoost algorithm. Taking the data of a province from 2019 to 2020 as an example, the proposed method is tested and verified. Using the basic information data, operation data and environmental data, the proposed method is compared with the traditional algorithms such as KNN, naive Bayes and support vector machine. The results show that the burning fault prediction of the XGBoost algorithm is better than the traditional algorithms. The precision of the XGBoost is 91%, the recall is 66%, and F1-score is 76.51%. In the process of system deployment, LSTM algorithm is used to fill some missing values. The experimental results show that the model can accurately predict the burning fault of smart meter in low-voltage platform area.